Hi all,

We all know that a univariate t-test for testing the equality of 2 means is valid if either

-          both samples come from normal distributions

-          when the samples are large such that  the sample means are normally distributed anyway because of the central limit theorem.

To generalize this test to the multivariate and multigroup case, I estimate a multivariate regression model with dummies as explanatory variables and use the Wilks Lambda teststatistic which is related to the eigenvalues of the inverse(Within covmatrix)*(Between covmatrix).  Now I am puzzled about the assumptions I need in this context.

Will large samples also make the Wilks Lambda test valid, even if the data do not come from multivariate normal distributions or do I need here multivariate normal distrubution of the data, no matter the size of the samples?

Please reply to me directly as it is not clear to me whether I am allowed to post this question here J…

I would be very grateful if someone could help me out…

Martina